长文自动摘要与自动封面：未填时从正文派生——摘要剥语法取开头 110 字（标题行/代码块/图链接剔除），封面取正文第一张图；更新只在「新旧都空」时补，站主主动清空的保得住；保存响应回写编辑器可见

2026-09-29 01:05:46 +08:00
parent 6e6196d68b
commit f7669badf2
4 changed files with 139 additions and 1 deletions
@@ -231,6 +231,7 @@ func (a *API) listPosts(w http.ResponseWriter, r *http.Request) {
 		}
 		in.Status = store.NormalizeStatus(in.Status)
 		attachLinkCard(r.Context(), &in, nil)
+		autoMeta(&in, nil)
 		p, err := a.Store.Create(in)
 		if err != nil {
 			httpx.ServerError(w, err)
@@ -270,12 +271,16 @@ func (a *API) postByID(w http.ResponseWriter, r *http.Request) {
 		if in.Status != "" {
 			in.Status = store.NormalizeStatus(in.Status)
 		}
-		// 多读一次当前行只为拿到已有卡片：链接没变就不必再打远端
+		// 多读一次当前行：链接卡片没变就不必再打远端；摘要/封面自动
+		// 生成也以它为准——「新值与旧值都为空」才补，主动清空的保得住
 		var curCard *model.LinkCard
+		var curPost *model.Post
 		if cur, err := a.Store.Get(id); err == nil {
 			curCard = cur.LinkCard
+			curPost = &cur
 		}
 		attachLinkCard(r.Context(), &in, curCard)
+		autoMeta(&in, curPost)
 		p, err := a.Store.Update(id, in)
 		writeOne(w, p, err)
 	case http.MethodDelete:
@@ -0,0 +1,68 @@
+// 长文的自动摘要与自动封面：站主没填这两项时从正文派生——
+// 摘要取正文开头的一段纯文本，封面取正文第一张图。
+// 更新路径只在「新值与旧值都为空」时才生成，站主主动清空的要保得住。
+package admin
+
+import (
+	"regexp"
+	"strings"
+
+	"oneblog/internal/model"
+)
+
+const autoSummaryLen = 110
+
+var (
+	mdImageRe   = regexp.MustCompile(`!\[[^\]]*\]\(([^)\s]+)[^)]*\)`)
+	mdCodeFence = regexp.MustCompile("(?s)```.*?```")
+	mdLinkRe    = regexp.MustCompile(`\[([^\]]*)\]\([^)]*\)`)
+	mdHeadingRe = regexp.MustCompile(`(?m)^#{1,6}\s+.*$`) // 标题行整行不进摘要（与文章标题重复）
+	mdNoiseRe   = regexp.MustCompile(`(?m)^([>\s*-]+|\d+\.\s)+`)
+	mdEmphasis  = strings.NewReplacer("**", "", "__", "", "~~", "", "*", "", "_", "")
+	wsRe        = regexp.MustCompile(`\s+`)
+)
+
+// autoMeta 按需填充 in.Summary / in.CoverURL。
+// cur 为更新前的旧文章（新建时传 nil）。
+func autoMeta(in *model.PostInput, cur *model.Post) {
+	if in.Kind != model.KindLong || strings.TrimSpace(in.ContentMd) == "" {
+		return
+	}
+	if strings.TrimSpace(in.Summary) == "" && (cur == nil || strings.TrimSpace(cur.Summary) == "") {
+		in.Summary = summarizeMarkdown(in.ContentMd, autoSummaryLen)
+	}
+	if strings.TrimSpace(in.CoverURL) == "" && (cur == nil || strings.TrimSpace(cur.CoverURL) == "") {
+		in.CoverURL = firstImage(in.ContentMd)
+	}
+}
+
+// summarizeMarkdown 剥掉 markdown 语法后取正文开头一段纯文本
+func summarizeMarkdown(md string, maxRunes int) string {
+	s := mdCodeFence.ReplaceAllString(md, " ") // 代码块整段不进摘要
+	s = mdHeadingRe.ReplaceAllString(s, " ")
+	s = mdImageRe.ReplaceAllString(s, " ")
+	s = mdLinkRe.ReplaceAllString(s, "$1")
+	s = mdNoiseRe.ReplaceAllString(s, " ")
+	s = mdEmphasis.Replace(s)
+	s = strings.ReplaceAll(s, "`", "")
+	s = wsRe.ReplaceAllString(s, " ")
+	s = strings.TrimSpace(s)
+	runes := []rune(s)
+	if len(runes) > maxRunes {
+		return string(runes[:maxRunes]) + "…"
+	}
+	return s
+}
+
+// firstImage 取正文第一张图片的地址（markdown 图；http/站内路径均可）
+func firstImage(md string) string {
+	m := mdImageRe.FindStringSubmatch(md)
+	if m == nil {
+		return ""
+	}
+	u := m[1]
+	if strings.HasPrefix(u, "http://") || strings.HasPrefix(u, "https://") || strings.HasPrefix(u, "/uploads/") {
+		return u
+	}
+	return ""
+}
@@ -0,0 +1,62 @@
+package admin
+
+import (
+	"strings"
+	"testing"
+
+	"oneblog/internal/model"
+)
+
+func TestSummarizeMarkdown(t *testing.T) {
+	md := "## 标题\n\n这是**正文**的第一段，含[链接](https://x.com)与图片：\n\n![封面](https://a.com/1.png)\n\n```go\ncode ignored\n```\n\n后续内容"
+	got := summarizeMarkdown(md, 110)
+	for _, bad := range []string{"#", "**", "![]", "```", "code ignored"} {
+		if strings.Contains(got, bad) {
+			t.Errorf("summary 含语法残留 %q: %q", bad, got)
+		}
+	}
+	if !strings.Contains(got, "链接") || !strings.HasPrefix(got, "这是") {
+		t.Errorf("summary 应以正文开头并保留链接文字: %q", got)
+	}
+	long := strings.Repeat("长", 200)
+	if got := summarizeMarkdown(long, 110); len([]rune(got)) != 111 || !strings.HasSuffix(got, "…") {
+		t.Errorf("超长应截到 110+省略号, got %d runes", len([]rune(got)))
+	}
+}
+
+func TestFirstImage(t *testing.T) {
+	md := "前言\n\n![封面](https://a.com/1.png)\n\n![第二张](https://a.com/2.png)"
+	if got := firstImage(md); got != "https://a.com/1.png" {
+		t.Errorf("firstImage = %q", got)
+	}
+	if got := firstImage("没有图片"); got != "" {
+		t.Errorf("无图应返回空, got %q", got)
+	}
+}
+
+func TestAutoMeta(t *testing.T) {
+	in := model.PostInput{Kind: model.KindLong, ContentMd: "# T\n\n正文内容 ![图](https://a.com/x.png)"}
+	autoMeta(&in, nil)
+	if in.Summary == "" || in.CoverURL != "https://a.com/x.png" {
+		t.Errorf("空字段应自动填充: summary=%q cover=%q", in.Summary, in.CoverURL)
+	}
+	// 站主已填的不能覆盖
+	in2 := model.PostInput{Kind: model.KindLong, ContentMd: "正文", Summary: "手写的", CoverURL: ""}
+	autoMeta(&in2, nil)
+	if in2.Summary != "手写的" {
+		t.Errorf("手写摘要不应被覆盖: %q", in2.Summary)
+	}
+	// 更新路径：新值与旧值都空才补——主动清空的保得住
+	cleared := model.PostInput{Kind: model.KindLong, ContentMd: "正文", Summary: "", CoverURL: ""}
+	cur := model.Post{Summary: "旧摘要", CoverURL: "旧封面"}
+	autoMeta(&cleared, &cur)
+	if cleared.Summary != "" || cleared.CoverURL != "" {
+		t.Errorf("主动清空不应复活: summary=%q cover=%q", cleared.Summary, cleared.CoverURL)
+	}
+	// 新旧都空 → 生成
+	never := model.PostInput{Kind: model.KindLong, ContentMd: "正文 ![图](/uploads/a.png)"}
+	autoMeta(&never, &model.Post{})
+	if never.CoverURL != "/uploads/a.png" {
+		t.Errorf("新旧都空应取首图: %q", never.CoverURL)
+	}
+}